Development of miniature climbing robots: Modeling, control and motion planning

An increasing interest in the development of special climbing robots has been witnessed in the last decade. Cleaning high-rise buildings, spray painting and sand blasting of gas tanks, inspecting and maintaining nuclear facilities, assisting fire fighting and rescue operations, remote monitoring of hazardous environment, etc.—all of those practical problems have an immediate need of automation. Climbing robots, with their ability to adhere to wall surfaces and move around carrying appropriate sensors or tools, are the best candidates for these kinds of jobs. The special characteristics and capabilities of climbing robots would not only allow them to replace human workers in these dangerous duties but also eliminate costly erection of scaffolding.; This dissertation describes the development of miniature bipedal climbing robots, with an emphasis on the CRAWLER robot. This robot is the smallest such robot to date, able to climb walls, walk on ceilings, travel through pipes, and transit between two inclined surfaces. Two active suction feet, where pump motors vacuum air out of suction cups, are used to support the robot on surfaces. The robot adopts the bipedal structure and an under-actuated mechanism to provide the robot with versatile mobility and multiple locomotion modes. The under-actuated mechanism reduces the number of motors required and thereby the robot size, weight, and power consumption, but it imposes challenges on robot control and motion planning.; The robot kinematic model and dynamic model are derived. The analysis of the dynamic model reveals that the robot link gravity is a dominant term in robot dynamic effects. A joint level PD (proportional + derivative) control plus feedforward gravity compensation method is proposed. This method outperforms the conventional PD control method because it not only achieves the joint level control but also compensates for the gravity effects which depend on the configuration of all the robot links. A DSP-based embedded control system is designed and installed on-board to control the robot. By using the DSP (digital signal processor) chip, the number of electrical components is minimized and the control system is self-contained.; In motion planning analysis, a hybrid configuration space concept is proposed which incorporates the continuous configuration space with the discrete motion status space i.e., standing foot, motion mode). The hybrid configuration space is an effective tool to analyze the motion planning of the climbing robot since the robot motion is uniquely determined in the hybrid configuration space framework. Based on motion pattern analysis, a motion planning method is developed that consists of a global planner and a local planner. The global planner generates a possible path by simplifying the robot as a rectangular rigid object with no kinematic constraints. By using an approach called trapezoidal decomposition and a searching algorithm known as A*, it is easy for the global planner to smooth the possible path and allow the robot to move effectively in translation mode. The possible path describes the global motion of the robot and minimizes the turns. The purpose of the local planner is to generate a feasible motion sequence around the turning point by considering the robot constraints. A cost function is defined based on the motion status information and a number of heuristics to help the search of an optimal motion sequence. This approach using global and local levels of refinement reduces the overall complexity and simplifies implementation. Experiments and simulations demonstrate the effectiveness of our robot system.